read_csv with names arg not implicitly setting header=None

[kieranholland]

>>> pandas.read_csv(StringIO('1,2'), names=['a', 'b'])
Empty DataFrame
Columns: [a, b]
Index: []
>>> pandas.read_csv(StringIO('1,2'), names=['a', 'b'], header=None)
   a  b
0  1  2

names is overloaded so that if header is not False, names means "choose these columns", rather like
the Dataframe constructor behaves.

sorry, I misread the docstring and expected it to be consistent with the dataframe ctor.

from the docstring:

names : array-like
    List of column names to use. If header=True, replace existing header,
    otherwise this ignores the header=0 default

when header=0 by default, the names argument is used to override the column names.
Not as consistent with the dataframe constructor, especially since header=False and header=0
are currently treated the same, so there's room to specify that behaviour.

[kieranholland]

Thanks for the explanation

My expectation came from the current development docs for read_table -

names: List of column names to use. If passed, header will be implicitly set to None.

I suspect overloading names is likely to cause confusion.

I see. The docs here match the docstring i quoted.
Can you provide a link to the docs you're refering to?

[kieranholland]

http://pandas.pydata.org/pandas-docs/dev/io.html

thanks!

looks like weirdness related to bool coercion, similar to #2199

In [36]: pd.read_csv(StringIO('1,2'), names=['a', 'b'],header=True)
---------------------------------------------------------------------------
CParserError                              Traceback (most recent call last)
<ipython-input-36-7af61abcf59b> in <module>()
----> 1 pd.read_csv(StringIO('1,2'), names=['a', 'b'],header=True)
/home/user1/src/pandas/pandas/_parser.so in pandas._parser.TextReader.__cinit__ (pandas/src/parser.c:3801)()

/home/user1/src/pandas/pandas/_parser.so in pandas._parser.TextReader._get_header (pandas/src/parser.c:4884)()

CParserError: Passed header=1 but only 1 lines in file

pd.read_csv(StringIO('1,2'), names=['a', 'b'],header=False)
Out[37]: 
Empty DataFrame
Columns: [a, b]
Index: []

pd.read_csv(StringIO('1,2'), names=['a', 'b'],header=0)
Out[38]: 
Empty DataFrame
Columns: [a, b]
Index: []

[wesm]

Note the API changes in the 0.10 release notes:

  - ``names`` handling in file parsing: if explicit column `names` passed,
    `header` argument will be respected. If there is an existing header column,
    this can rename the columns. To fix legacy code, put ``header=None`` when
    passing ``names``

It's unfortunate but automatically forcing header=None was making handling some files (in which renaming columns was necessary) impossible.

We need to make a sweep through the docs to reflect the API change.